EDP Sciences
Free Access
Issue
A&A
Volume 616, August 2018
Article Number L4
Number of page(s) 7
Section Letters to the Editor
DOI https://doi.org/10.1051/0004-6361/201833418
Published online 14 August 2018

© ESO 2018

1. Introduction

Metallicity provides one of the most important clues in understanding galaxy formation and evolution since metal formation and enrichment are closely connected with the past star formation history in galaxies. Generally, the metallicity of star-forming galaxies is obtained by using various emission lines at the rest-frame optical wavelengths (e.g., [O II]λ3727, Hβ, [O III]λ5007, Hα, and [N II]λ6584), emitted in H ii regions (e.g., Nagao et al. 2006a; Curti et al. 2017). The relationship between mass of galaxies and their metal (the so-called mass-metallicity relation, MZR) has been investigated for about forty years (Lequeux et al. 1979). By using Sloan Digital Sky Survey (SDSS) imaging and spectroscopy of ~53 000 star-forming galaxies at z ~ 0.1, Tremonti et al. (2004) made a major step forward in this field by obtaining a highly statistically sound relationship between stellar mass and gas-phase oxygen abundances in galaxies (see also Kewley & Ellison 2008; Hayashi et al. 2009; Andrews & Martini 2013). Furthermore, this correlation has been explored at high redshifts, up to z ~ 3, and a clear evolution with cosmological time, has been reported (e.g., Savaglio et al. 2005; Maiolino et al. 2008; Mannucci et al. 2009; Cresci et al. 2012; Yuan et al. 2013; Hunt et al. 2016; Onodera et al. 2016; Suzuki et al. 2017).

The metallicity of star-forming galaxies at high redshift has been estimated through deep near-infrared spectroscopy. However, since the emission lines for the metallicity diagnostics shift out of the K-band atmospheric window for galaxies at z > 3.5, studies are more challenging for higher-z galaxies requiring space-based spectrograph. An alternative approach to investigate the metallicity in the early universe is to focus on active galactic nuclei (AGNs) instead of star-forming galaxies. Thanks to their high luminosity and strong emission lines in the wide wavelength range from the ultraviolet (UV) to the infrared, arising from gas clouds photoionized by their central accreting black hole, it is possible to measure the metallicity even in high-z universe. Since Hamann & Ferland (1992) proposed that broad-line region (BLR) metallicities (ZBLR) can be estimated from N Vλ1240, C IVλ1549, and He IIλ1640 lines (see also Hamann & Ferland 1993, 1999), AGN metallicities have been studied extensively. By measuring emission-line flux ratios of SDSS quasars, i.e., N Vλ1240/C IVλ1549, (Si IVλ1398+O IVλ1402)/C IVλ1549, Al IIIλ1857/C IVλ1549, and N Vλ1240/He IIλ1640, Nagao et al. (2006b) investigated BLR metallicities and their properties. Based on a spectral stacking analysis, they found a positive correlation between BLR metallicities and AGN luminosities (LAGN): more luminous quasars show more metal-rich BLR clouds (see also Hamann & Ferland 1993; Juarez et al. 2009). By using SDSS spectra of 2383 quasars at 2.3 < z < 3.0, Matsuoka et al. (2011a) revealed that the observed LAGNZBLR trend is mostly a consequence of the relation between black-hole mass (MBH) and BLR metallicity. This result indicates that AGN host galaxies may follow a MZR given the relation between black-hole mass and galaxy mass. However, it is unclear if the BLR metallicities trace chemical properties of their host galaxies, since AGN broad lines originate from a very small region in galactic nuclei, RBLR < 1 pc (e.g., Suganuma et al. 2006), which may have evolved more rapidly and may not be representative of the metallicity in their host galaxies.

To investigate the metallicity on the scale of AGN host galaxies, the narrow line region (NLR) has been studied, since the typical NLR size is comparable to the size of their host galaxies, RNLR ~ 101−4 pc (e.g., Bennert et al. 2006b,a). Furthermore, NLR masses are usually ~104−8 M, much larger than those of the BLR, ~102−4 M (e.g., Baldwin et al. 2003). Thus, NLR metallicities allow us to investigate chemical properties at host galaxy scales. However, since it has been difficult to discover type-2 quasars in the high-z universe, the NLR metallicity at z > 1 have been investigated only with high-z radio galaxies (HzRGs) in previous studies. Nagao et al. (2006c) studied NLR metallicities of 51 HzRGs at 1.2 < z < 3.8 with a metallicity diagnostic diagram involving C IVλ1549, He IIλ1640, and C IIIλ1909. They reported that HzRGs do not show any redshift evolution of NLR metallicity: a lack of redshift evolution is similar to that seen for the BLR metallicities (Nagao et al. 2006b; Juarez et al. 2009). Matsuoka et al. (2009) confirmed the absence of any significant NLR metallicity evolution up to z ~ 4 by observing nine HzRGs at z > 2.7. Moreover, Matsuoka et al. (2011b) found that the most distant radio galaxy, TN J0924−2201 at z = 5.19, has already experienced significant chemical evolution.

On the other hand, Matsuoka et al. (2009) also found a positive correlation between NLR metallicities and AGN luminosities, as for the BLR metallicities. This may be originated from the galaxy mass-metallicity relation, if we assume LAGNMBHMhost. However, there is still no direct evidence that NLR metallicities are related to their host properties, and in particular with their stellar mass (M). In this Letter, we focus on HzRGs and X-ray selected type-2 AGNs at 1.2 < z < 4.0 and investigate their NLR metallicities, adopting a diagnostic diagram of C IVλ1549, He IIλ1640, and C IIIλ1909 emission lines. Thus, by collecting their stellar masses from the recent literature we examine their dependence on NLR metallicities.

2. Observations and Data from the Literature

In order to study NLR metallicities, we utilize type-2 AGNs, i.e. in which the BLR and strong continuum emission are obscured allowing us accurate measurements of narrow-line emission. First, we focus on HzRG samples (e.g., De Breuck et al. 2000) which have been identified as type-2 AGNs at high redshift with radio observations: e.g., ultra-steep-spectrum radio sources with radio spectral index α < −1.30 (Svvα) are adopted as efficient HzRG tracers in De Breuck et al. (2000) and over 160 sources are identified as HzRGs at 0 < z < 5.2 with devoted optical spectroscopic observations. Requiring that C IVλ1549, He IIλ1640, and C IIIλ1909 lines are detected, 67 HzRGs at 1.2 < z < 4.0 are collected from Nagao et al. (2006c), Bornancini et al. (2007), and Matsuoka et al. (2009, 2011a). We listed these emission-line fluxes in Appendix A.

As inferred from the tight correlation in the Hubble K-z diagram (e.g., De Breuck et al. 2002), HzRGs are usually associated with very massive elliptical galaxies (M > 1011M; e.g., Seymour et al. 2007). In order to also sample the lower stellar mass range, in this study we also focus on X-ray selected type-2 AGNs which have been discovered in recent observations. First, we obtained optical spectra of two X-ray selected radio-quiet type-2 AGNs at z > 3.0 in the Cosmic Evolution Survey (COSMOS) field (Mainieri et al. 2011) with FOCAS (the Faint Object Camera And Spectrograph; Kashikawa et al. 2002) at the Subaru Telescope (24−25 December 2013). Observations were performed with the 300R dispersion element and SY47 filter to cover the wavelength range of λobs = 4900−9500 Å. We used a slit width of 0″̣6. Standard data reduction procedures were applied to these spectra by using IRAF tasks. Figure 1 shows the final reduced spectra. We measured emission-line fluxes of detected lines by fitting them with a Gaussian function using the IRAF task splot (see Appendix A). For undetected lines, 3σ upper limits are given by assuming an averaged line width of 25 Å calculated from the detected lines. Then, we collected emission-line fluxes of 10 X-ray selected AGNs at 1.5 < z < 4.0 from Nagao et al. (2006c). Nine of them are X-ray selected type-2 AGNs in the Chandra Deep Field-South (CDF-S) field (Szokoly et al. 2004) and the remaining CXO 52 is a type-2 quasar in the Lynx field identified as a hard X-ray source with Chandra (Stern et al. 2002). Emission line fluxes of these 13 X-ray type-2 AGNs are listed in Appendix A.

thumbnail Fig. 1.

Reduced spectra of two X-ray selected type-2 AGNs in the COSMOS field (ID: 05162 and 10690) obtained with FOCAS, after adopting 4 pixel binning in wavelength direction. Their two-dimensional spectra with a 2″̣1 section are shown at the top, respectively. Red vertical lines indicate the central wavelengths of rest-frame UV emission lines, i.e., Lyλ, N Vλ1240, C IVλ1549, He IIλ1640, and C IIIλ1909. The gray line in the top panel is an arbitrary spectrum with five times the flux for recognising relatively weak emission lines.

Open with DEXTER

We collected their stellar masses from the literature. Regarding HzRGs, we matched our sample with stellar mass catalogs in Seymour et al. (2007) and De Breuck et al. (2010). Overall, we obtained stellar masses of 21 HzRGs including 8 upper limits. These studies derived stellar masses through a spectral energy distribution (SED) fit based on the Spitzer infrared (3.6–160 μm) data, covering the rest-frame optical to infrared wavelengths (see Seymour et al. 2007, for more detail): De Breuck et al. (2010) followed the procedures of Seymour et al. (2007) with the same method and photometric bands. For X-ray selected AGNs in the COSMOS field, stellar masses are collected from Mainieri et al. (2011). These masses are also derived by SED fitting with multiwavelength data set of COSMOS. For five additional AGNs in CDF-S, we adopted stellar masses derived using a tight correlation between rest-frame optical colors and stellar mass-to-light ratios in Xue et al. (2010). Altogether, we obtained stellar masses of 28 type-2 AGNs containing 8 HzRGs with upper limits (see Appendix A).

3. Result

We calculate emission-line flux ratios of C IVλ1549/He IIλ1640 and C IIIλ1909/C IVλ1549, that are sensitive to both the NLR gas metallicity and ionization parameter (see Nagao et al. 2006c; Matsuoka et al. 2009). Figure 2 shows our sample plotted on a diagnostic diagram of C IVλ1549, He IIλ1640, and C IIIλ1909 lines in the left panel. Our new type-2 AGNs are located in the same area as the X-ray selected AGNs in Nagao et al. (2006c). In the right panel, 28 type-2 AGNs for which stellar masses were derived are plotted, color-coded according to their masses: we gave the stellar mass distribution as a function of redshift in the middle panel. Thanks to X-ray selected type-2 AGNs, we could cover the low-M range, i.e., log(M/M) < 11, which could not be explored with HzRGs.

thumbnail Fig. 2.

Emission-line flux ratios of high-z type-2 AGNs plotted on the diagnostic diagram of C IVλ1549/He IIλ1640 versus C IIIλ1909/C IVλ1549 (left). Gray- and black-filled squares and arrows denote HzRGs compiled by Nagao et al. (2006c) and Matsuoka et al. (2009), respectively. The C IVλ1549/He IIλ1640 lower limit for TN J0924−2201 is shown as a light-gray arrow (Matsuoka et al. 2011a). Our new type-2 AGNs are shown as red-filled double circles (COSMOS 05162 and 10690). Orange circles denote X-ray selected type-2 AGNs in Nagao et al. (2006c). Middle panel: objects in our sample having an estimate of stellar mass as a function of redshift, and their emission-line flux ratios are plotted in the right panel. In the middle and right panels, symbols are same as the left panel but color-coded according to the stellar mass: bluer and redder for lower and higher mass objects, respectively. Note that open squares in these panels indicate HzRGs with upper limits of stellar masses.

Open with DEXTER

4. Discussion

In order to examine the relation between NLR metallicities and stellar masses statistically, we divided the 20 type-2 AGNs whose stellar masses are estimated without upper limits into three subsamples with stellar mass ranges, i.e., log(M/M) < 11.0, 11.0 < log(M/M) < 11.5, and log(M/M) > 11.5 (these intervals are shown with horizontal lines in the middle panel in Fig. 2). Then, we calculated averaged emission-line flux ratios of C IVλ1549/He IIλ1640 and C IIIλ1909/C IVλ1549 for each subsample: we use 15 objects whose emission-line flux ratios are measured without upper or lower limits (i.e., three, nine, and three objects, respectively). Note that our final sample would not be biased due to the discard of objects with upper or lower limits since the distribution of them show no obvious difference from the original sample (see the right panel of Fig. 2). Figure 3 shows these averaged flux ratios on the diagnostic diagrams (top panels). Using these averaged points, we estimate NLR metallicities for each stellar mass.

thumbnail Fig. 3.

Top panels: Averaged emission-line flux ratios of low (blue), middle (purple), and high (red) mass subsamples plotted on the diagnostic diagram on C IVλ1549, He IIλ1640 and C IIIλ1909. The model predictions with different gas densities (log nH = 1, 4, and 5) are presented in the left, middle, and right panels, respectively. Constant metallicity and constant ionization parameter sequences are denoted by orange and blue solid lines, respectively. The gray bars denote the root mean square of the data distribution and the estimated errors in the averaged values are shown as color bars. Bottom panels: NLR metallicities as a function of stellar masses, estimated with different three gas density models. Open circles would be less-likely solutions (see text).

Open with DEXTER

To estimate NLR metallicities from observational emissionline flux ratios, we carried out model calculations using the photoionization code Cloudy (version 17.00; Ferland et al. 1998, 2017): NLR clouds are mainly photoionized and not significantly affected by shocks when we focus on UV emission lines (see Matsuoka et al. 2009). We assumed three cloud models with hydrogen densities of nH = 101, 104, and 105 cm−3. For each model, we calculated emission-line flux ratios by spanning ionization parameter and metallicity in the following ranges U = 10−3.0−10−0.4 and ZNLR = 0.1−10Z, with step values of Δ log U = 0.2 and Δ log(ZNLR/Z) = 0.01, respectively (i.e., 8442 models in total). We assume a typical AGN SED as photoionizing continuum radiation, using the “table AGN” command in Cloudy (Mathews & Ferland 1987). Chemical compositions of NLR gas clouds are scaled by maintaining solar abundance ratios except for helium and nitrogen. We adopted analytical expressions for He and N relative abundances as a function of metallicities (see Dopita et al. 2000; Russell & Dopita 1992; van Zee et al. 1998). We stopped calculations when the hydrogen ionization fraction dropped below 15%.

Our model predictions of C IVλ1549/He IIλ1640 and C IIIλ1909/C IVλ1549 are overplotted on the diagnostic diagrams (top panels of Fig. 3): metallicity-constant and U-constant sequences are denoted as orange and blue lines, respectively. Our models confirm these photoionization models are useful to investigate NLR metallicities, although they show some dependences on hydrogen gas densities. Comparing these model predictions with our M-averaged line ratios we estimated mean NLR metallicities. We show results of ZNLR estimates as a function of stellar masses for each nH model in bottom panels of Fig. 3. We estimated NLR metallicities for each M-averaged line ratio and their errors from averaged emission-line flux ratios and their error bars, respectively, matching them to model calculations with a metallicity step of log(ZNLR/Z) = 0.01. Since absolute NLR metallicities are not determined uniquely due to dependences of gas densities, we compare the relative ZNLR differences for each density. This approach should be reliable since Sanders et al. (2016) reported no dependence of gas density on stellar mass investigating the [O II]λλ3726,3729 and/or [S II]λλ6716,6731 doublets of local and high-z starforming galaxies at z ~ 2.3 (see also Shimakawa et al. 2015; Kaasinen et al. 2017). Note that in principle we derived two values of metallicities if emission-line flux ratios are located in a degenerated area on the model predictions. However, lower metallicity in the middle-M bin is unplausible as a significant decrease of metallicity around medium-masses has never been seen for any population of galaxies, at any redshift, and not conceived by any model or cosmological simulation, therefore we exclude such lower metallicities solution in the middle-Mbin, and consider only the upper solution for this bin.

As shown in these panels of Fig. 3, we find positive correlations between NLR metallicities and stellar masses in type-2 AGNs at 1.2 < z < 4.0 at different fixed densities. This result is the first indication that AGN gas metallicities are related to their host properties, i.e., host galaxy masses in this case, most likely indicating ZNLRZgal if we assume that AGN host galaxies have the same MZR in the star-forming galaxies. We thus propose that NLR metallicity is correlated with host metallicity, although additional data sets would be needed to confirm this statistically.

The same averaged emission-line flux ratios shown in top panels of Fig. 3 may also be reproduced if there is an anti-correlation between stellar masses and densities instead of the MZR. However, some observational studies of star-forming galaxies at 0 < z < 2.5 indicated that electron densities do not show any such dependence on stellar masses. For example, Kaasinen et al. (2017) found no significant Mne trend of star-forming galaxies at z ~ 1.5 although local SDSS galaxies indicate a weak positive correlation between electron density and stellar mass, indicating the opposite sense. Moreover, Shimakawa et al. (2015) reported any correlation between Mand ne cannot be identified with H -selected star-forming galaxies at z = 2.5 (see also Sanders et al. 2016). Note here that we assume these Mne relation of star-forming galaxies are applicable to NLR densities. Therefore, we conclude that our mass-averaged emission-line flux ratios can be explained mainly by the dependence on NLR metallicities.

Although there are uncertainties associated with density dependences, we compare our MZR of type-2 AGNs with those of galaxies. Figure 4 shows the redshift evolution of the MZRs for star-forming galaxies at z = 0.07, 0.7, 2.2, and 3.1 (Kewley & Ellison 2008; Savaglio et al. 2005; Erb et al. 2006; Mannucci et al. 2009), with our results are overplotted. Since the mean redshift of our sample is ~2.7, we focus on the MZR of star-forming galaxies at z = 3.1 as a comparison. As shown in this figure, we indicate that MZRs obtained with high-nH models (nH = 104 and 105 cm3 shown as an orange shaded area) seems consistent with MZR of star-forming galaxies. Recently, Nitta et al. (in prep.) has found that NLR gas densities of high-z AGNs are relatively higher than those of low-z ones. By investigating restframe optical emission lines of quasars, they estimated a typical gas density of ~105.5 cm−3 (see also Araki et al. 2012).

thumbnail Fig. 4.

The mass-metallicity relation of type-2 AGNs at z ~ 3 comparing with MZR of star-forming galaxies. Gray solid curves are MZR of star-forming galaxies at z = 0.07, 0.7, and 2.2, and red curve is for z = 3.1 galaxies in Mannucci et al. (2009). Filled and open circles are same as Fig. 3, and the orange area denotes MZR of type-2 AGNs which NLR metallicities are estimated with high-nH models.

Open with DEXTER

Note that some studies (e.g., Nagao et al. 2006c; Matsuoka et al. 2009) about the chemical evolution of HzRGs claimed that there is no significant metallicity evolution in NLRs, up to z ~ 4, apparently contradicting our result that HzRGs follow the redshift-dependent MZR of star-forming galaxies. However, a possibility is that it is not possible to identify such redshift evolution only focusing on HzRGs hosted in massive elliptical galaxies. Probably, the inclusion of X-ray selected AGNs with log(M/M) ~ 10.5 has helped to reveal such redshift-evolving MZR in type-2 AGNs as shown in this study. Moreover, star formation rate (SFR) or gas mass should be crucial parameters to include, as recent studies have revealed that the relation between stellar mass, gas-phase metallicity, and SFR or gas mass (fundamental metallicity relation: FMR) is important to understand the chemical evolution of galaxies (e.g., Mannucci et al. 2010; Lara-López et al. 2010; Bothwell et al. 2016). To investigate the FMR of type-2 AGNs statistically we need a larger sample and also access to SFR diagnostics that are not affected by the AGN activity, e.g., the far-infrared luminosity (e.g., Beelen et al. 2006; Matsuoka & Woo 2015; Matsuoka & Ueda 2017).

5. Conclusion

We have investigated the relation between NLR metallicities and stellar masses of type-2 AGNs at 1.2 < z < 4.0 by using a pilot sample of 11 HzRGs and four X-ray selected type-2 AGNs and found the following results:

  1. We found indications of a positive correlation between NLR metallicities and stellar masses of type-2 AGNs at 1.2 < z < 4.0; this is the first direct evidence that AGN metallicities are related to their host properties.

  2. Although there are uncertainties associated with the density dependence of the metallicity diagnostic diagrams, we have found that the mass-metallicity relation of type-2 AGNs, as inferred by us, are compatible with that of star-forming galaxies at similar redshifts, indicating that NLR metallicity is connected with host metallicities.

Acknowledgments

We would like to thank Takashi Hattori for assisting our FOCAS observations (S13B-019 and S14B-003). We are also grateful to Gray J. Ferland for providing the great photoionization code Cloudy. Data analysis were in part carried out on common-use data analysis computer system at the Astronomy Data Center, ADC, of the National Astronomical Observatory of Japan (NAOJ). K.M. is supported by Japan Society for the Promotion of Science (JSPS) Overseas Research Fellowships. T.N. is supported by JSPS KAKENHI Grant Nos. 16H01101, 16H03958, and 17H01114. R.M. acknowledges support by the Science and Technology Facilities Council (STFC) and the ERC Advanced Grant 695671 “QUENCH”.

References

Appendix A

Table A.1.

Emission-line measurements of type-2 AGNs and their host stellar masses.

All Tables

Table A.1.

Emission-line measurements of type-2 AGNs and their host stellar masses.

All Figures

thumbnail Fig. 1.

Reduced spectra of two X-ray selected type-2 AGNs in the COSMOS field (ID: 05162 and 10690) obtained with FOCAS, after adopting 4 pixel binning in wavelength direction. Their two-dimensional spectra with a 2″̣1 section are shown at the top, respectively. Red vertical lines indicate the central wavelengths of rest-frame UV emission lines, i.e., Lyλ, N Vλ1240, C IVλ1549, He IIλ1640, and C IIIλ1909. The gray line in the top panel is an arbitrary spectrum with five times the flux for recognising relatively weak emission lines.

Open with DEXTER
In the text
thumbnail Fig. 2.

Emission-line flux ratios of high-z type-2 AGNs plotted on the diagnostic diagram of C IVλ1549/He IIλ1640 versus C IIIλ1909/C IVλ1549 (left). Gray- and black-filled squares and arrows denote HzRGs compiled by Nagao et al. (2006c) and Matsuoka et al. (2009), respectively. The C IVλ1549/He IIλ1640 lower limit for TN J0924−2201 is shown as a light-gray arrow (Matsuoka et al. 2011a). Our new type-2 AGNs are shown as red-filled double circles (COSMOS 05162 and 10690). Orange circles denote X-ray selected type-2 AGNs in Nagao et al. (2006c). Middle panel: objects in our sample having an estimate of stellar mass as a function of redshift, and their emission-line flux ratios are plotted in the right panel. In the middle and right panels, symbols are same as the left panel but color-coded according to the stellar mass: bluer and redder for lower and higher mass objects, respectively. Note that open squares in these panels indicate HzRGs with upper limits of stellar masses.

Open with DEXTER
In the text
thumbnail Fig. 3.

Top panels: Averaged emission-line flux ratios of low (blue), middle (purple), and high (red) mass subsamples plotted on the diagnostic diagram on C IVλ1549, He IIλ1640 and C IIIλ1909. The model predictions with different gas densities (log nH = 1, 4, and 5) are presented in the left, middle, and right panels, respectively. Constant metallicity and constant ionization parameter sequences are denoted by orange and blue solid lines, respectively. The gray bars denote the root mean square of the data distribution and the estimated errors in the averaged values are shown as color bars. Bottom panels: NLR metallicities as a function of stellar masses, estimated with different three gas density models. Open circles would be less-likely solutions (see text).

Open with DEXTER
In the text
thumbnail Fig. 4.

The mass-metallicity relation of type-2 AGNs at z ~ 3 comparing with MZR of star-forming galaxies. Gray solid curves are MZR of star-forming galaxies at z = 0.07, 0.7, and 2.2, and red curve is for z = 3.1 galaxies in Mannucci et al. (2009). Filled and open circles are same as Fig. 3, and the orange area denotes MZR of type-2 AGNs which NLR metallicities are estimated with high-nH models.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.